Published research articles in Machine Learning, AI and Data Science. More technical work and open-source contributions are on my GitHub.
Peer-Reviewed Publications
Peer-Reviewed IEEE Digital Xplore LibraryAccepted · September 2024
Predictive Maintenance in Automotive using Machine Learning
This research applies advanced ML techniques for predictive maintenance in automotive systems to minimize unplanned
downtime. High-frequency time-series sensor data from critical components like engines and brakes are analyzed using
supervised learning models - Random Forest, SVM and LSTM. Data preprocessing (noise reduction and feature engineering)
improves accuracy, while Grid Search and Cross-Validation tune hyperparameters. Feature importance is assessed via SHAP
for interpretability.
The study uses Scikit-learn, TensorFlow and PySpark for training, with distributed computing on Apache Hadoop and Spark
for large-scale data. Models are deployed on AWS Lambda for real-time failure prediction - demonstrating reduced
failure risk, improved reliability and lower maintenance costs. Accepted for publication in the IEEE Journal,
February 2025.
Peer-Reviewed Int'l Journal of Applied Engineering & TechnologyDecember 2023
Optimizing Supply Chain Management Through Data Science and AI: A Data-Driven Approach
This research investigates the optimization of supply chain processes through advanced data analytics and machine
learning. We focus on enhancing efficiency and resilience by employing predictive analytics over large-scale datasets,
identifying operational bottlenecks and forecasting demand variability. The framework integrates statistical process
control, ML algorithms and real-time data processing using Python, R and SQL.
A hybrid approach combining descriptive and prescriptive analytics evaluates supply chain performance - lead time,
inventory turnover and order fulfillment. Case studies from manufacturing and logistics demonstrate the efficacy of
data-driven decision-making, underscoring the importance of data integration, ML models and simulation in achieving
supply-chain agility.
Streamlining Data Engineering and ML Workflows with AWS
A practical guide to optimizing data engineering and machine learning workflows using AWS services. This article covers tools and best practices for building scalable data pipelines, automating ML processes, and enhancing productivity in data-driven projects.
Covers Lambda, Glue, Redshift, and integration patterns for production systems.
Revolutionizing Road Safety: How Machine Learning is Reducing Traffic Accidents
Explores how machine learning and predictive analytics are transforming road safety by analyzing driver behavior patterns and accident risk factors.
Covers real-world applications of ML in transportation, incident prediction systems, and measurable safety improvements in fleet operations.
Framework Wars Are Cheaper Than Most People Think: Benchmarking MCP, OpenAI Function Calling & LangGraph
A detailed 6-week benchmark study comparing three major agent orchestration frameworks - Model Context Protocol (MCP), OpenAI function calling, and LangGraph - on the same backend.
Key finding: framework choice moves roughly 200 tokens per request, but tool definition overhead sits at 1,100+ tokens regardless of orchestration library.
Demonstrates that selective tool exposure is the actual lever for cost optimization.
Conducted as part of Google Summer of Code 2026 with the Accord Project, mentored by Niall Roche and Dan Selman. Includes full technical writeup and implementation guidance for production systems.
The Hardest Part of Building an RL Environment Isn't the Environment - It's the Grader
A technical deep dive into reinforcement learning environment engineering, exploring a common but overlooked challenge: designing robust evaluation and grading systems.
Examines why environment implementations often work reliably while graders (the evaluation logic) become the source of failures, and provides patterns for addressing this asymmetry.